Recommender System Based On Trendsetter Inference

Abstract

A trend setting score that identifies a degree of trend setting exhibited by a user is generated for each of multiple users. This degree of trend setting exhibited by the user is an indication of how well the user identifies trends for items (e.g., consumes items) prior to the items becoming popular. The item consumption of users with high trend setting scores is then used to identify items that are expected to become popular after a lag in time. For a given user, another user with a high trend setting score (also referred to as a trendsetter) and having a high affinity with (e.g., similar item consumption behavior or characteristics) the given user is identified. Recommendations are provided to the given user based on items consumed by the trendsetter prior to the items becoming popular.

What is claimed is: 
     
         1 . In a digital medium environment to cause an item recommendation to be delivered to a client device, a method implemented by at least one computing device, the method comprising:
 generating, by the at least one computing device and for each of multiple users, a trend setting score identifying a degree of trend setting exhibited by the user by identifying a frequency with which the user consumes items that gain popularity after an amount of time, wherein whether an item consumed by the user gains popularity after the amount of time is determined by providing item consumption data for the item to a machine learning system trained to determine based on the item consumption data whether the item became popular the amount of time after the user consumed the item;   determining, by the at least one computing device for a first user of the multiple users, a second user of the multiple users having characteristic or behavior representations within a threshold distance of characteristic or behavior representations of the first user;   identifying, by the at least one computing device, a first item that is not currently popular but has been consumed by the second user; and   causing, by the at least one computing device, a recommendation for the first item to be delivered to the first user.   
     
     
         2 . The method as recited in  claim 1 , the machine learning system having been trained using training data including, for multiple items, item consumption data for a duration of time and along with a tag indicating whether the item consumption data indicates that the item became popular an amount of time after a user consumed the item. 
     
     
         3 . The method as recited in  claim 1 , the generating the trend setting score further comprising:
 determining a number of first additional users of the multiple users that consumed the item after the item became popular; and   incorporating the number of first additional users into the trend setting score.   
     
     
         4 . The method as recited in  claim 3 , the generating the trend setting score further comprising:
 determining trend setting scores of second additional users of the multiple users that consumed the item after the amount of time; and   incorporating the trend setting scores of the second additional users into the trend setting score.   
     
     
         5 . The method as recited in  claim 4 , the generating the trend setting score further comprising:
 determining activity of the user on social media;   assigning a social media score to the user based on the activity; and   incorporating the social media score into the trend setting score.   
     
     
         6 . The method as recited in  claim 5 , the generating the trend setting score further comprising generating, as the trend setting score, a value by summing:
 the frequency with which the user consumes items that subsequently gain popularity after the amount of time weighted by a first weight;   the number of first additional users of the multiple users that consumed the item after the item became popular weighted by a second weight;   the social media score weighted by a third weight; and   the trend setting scores of second additional users of the multiple users that consumed the item after the amount of time weighted by a fourth weight.   
     
     
         7 . The method as recited in  claim 1 , the generating the trend setting score for a user comprising:
 determining activity of the user on social media;   assigning a social media score to the user based on the activity; and   incorporating the social media score into the trend setting score.   
     
     
         8 . The method as recited in  claim 1 , the generating the trend setting score for a user comprising:
 determining trend setting scores of additional users of the multiple users that consumed the item after the amount in time; and   incorporating the trend setting scores of the additional users into the trend setting score.   
     
     
         9 . A system comprising:
 a user scoring module implemented at least partially in hardware of at least one computing device and configured to generate, for each of multiple users, a trend setting score identifying a degree of trend setting exhibited by the user by identifying a frequency with which the user consumes items that gain popularity after an amount of time, wherein whether an item consumed by the user gains popularity after the amount of time is determined by providing item consumption data for the item to a machine learning system trained to determine based on the item consumption data whether the item became popular the amount of time after the user consumed the item;   a user affinity module implemented at least partially in hardware of the at least one computing device and configured to determine, for a first user of the multiple users, a second user of the multiple users having characteristic or behavior representations within a threshold distance of characteristic or behavior representations of the first user; and   a trendsetter behavior identification module implemented at least partially in hardware of the at least one computing device and configured to identify a first item that is not popular but has been consumed by the second user for delivery of a recommendation for the first item to the first user via a recommendation delivery platform.   
     
     
         10 . The system as recited in  claim 9 , the user scoring module further generating the trend setting score by:
 determining a number of additional users of the multiple users that consumed the item after the item became popular; and   incorporating the number of additional users into the trend setting score.   
     
     
         11 . The system as recited in  claim 10 , the user scoring module incorporating the number of additional users into the trend setting score by generating, as the trend setting score, a value by combining the frequency with which the user consumes items that subsequently gain popularity after the amount of time weighted by a first weight and the number of additional users of the multiple users that consumed the item after the item became popular weighted by a second weight. 
     
     
         12 . The system as recited in  claim 9 , the user scoring module further generating the trend setting score by:
 determining trend setting scores of additional users of the multiple users that consumed the item after the amount of time; and   incorporating the trend setting scores of the additional users into the trend setting score.   
     
     
         13 . The system as recited in  claim 9 , the user scoring module further generating the trend setting score by:
 determining activity of the user on social media;   assigning a social media score to the user based on the activity; and   incorporating the social media score into the trend setting score.   
     
     
         14 . In a digital medium environment to cause an item recommendation to be delivered to a client device, a method implemented by at least one computing device, the method comprising:
 identifying, by the at least one computing device for a first user of multiple users, a set of users including two or more of the multiple users having characteristic or behavior representations within a threshold distance of characteristic or behavior representations of the first user;   generating, by the at least one computing device and for each user of the set of users, a trend setting score identifying a degree of trend setting exhibited by the user from one or more factors including a frequency with which the user consumes items that gain popularity after an amount of time, a number of first additional users of the multiple users that consumed items after the amount of time, a social media score, and trend setting scores of second additional users of the multiple users that exhibit similar trends after the amount of time, wherein items that gain popularity after an amount of time are items for which a rule indicates the items are popular;   determining, by the at least one computing device, a second user of the set of users that has a highest trend setting score of the set of users;   identifying, by the at least one computing device, a first item that is not currently popular but has been consumed by the second user; and   causing, by the at least one computing device, a recommendation for the first item to be communicated over a network for display to the first user.   
     
     
         15 . The method as recited in  claim 14 , the generating the trend setting score further comprising:
 determining a number of first additional users of the multiple users that consumed the item after the item became popular; and   incorporating the number of first additional users into the trend setting score.   
     
     
         16 . The method as recited in  claim 15 , the generating the trend setting score further comprising:
 determining trend setting scores of second additional users of the multiple users that consumed the item after the amount of time; and   incorporating the trend setting scores of the second additional users into the trend setting score.   
     
     
         17 . The method as recited in  claim 16 , the generating the trend setting score further comprising:
 determining activity of the user on social media;   assigning a social media score to the user based on the activity; and   incorporating the social media score into the trend setting score.   
     
     
         18 . The method as recited in  claim 17 , the generating the trend setting score further comprising generating, as the trend setting score, a value by summing:
 the frequency with which the user consumes items that subsequently gain popularity after the amount of time weighted by a first weight;   the number of first additional users of the multiple users that consumed the item after the item became popular weighted by a second weight;   the social media score weighted by a third weight; and   the trend setting scores of second additional users of the multiple users that consumed the item after the amount of time weighted by a fourth weight.   
     
     
         19 . The method as recited in  claim 14 , the generating the trend setting score for a user comprising:
 determining activity of the user on social media;   assigning a social media score to the user based on the activity; and   incorporating the social media score into the trend setting score.   
     
     
         20 . The method as recited in  claim 14 , the generating the trend setting score for a user comprising:
 determining trend setting scores of additional users of the multiple users that consumed the item after the amount in time; and   incorporating the trend setting scores of the additional users into the trend setting score.